Hardware

GeForce RTX 4070 Super

GeForce RTX 4070 Super has 12 GB GDDR6X and 504 GB/s. See which open models fit, at which quant, and roughly how fast. Figures below use the default memory (12 GB), 32 GB system RAM, and DDR5-5600 for anything that spills. Updated 2026-10-17.

Memory12 GB GDDR6X

Bandwidth504 GB/s

Kindgpu

Year2024

NVIDIA GeForce RTX 40-series specifications. Bandwidth is the published peak. Source.

RSS for new models on this device

Models that fit

Model	Verdict	Quant	Context	tok/s
Qwen3 14B 14.8B	Tight fit	Q5_K_S	4K	25–35
NousResearch Hermes 4 14B 14.8B	Tight fit	Q5_K_S	4K	25–35
Ministral 3 14B Instruct 2512 13.5B	Tight fit	UD-Q5_K_XL	8K	25–34
Ministral 3 14B Reasoning 2512 13.5B	Tight fit	UD-Q5_K_XL	8K	25–34
Qwen2.5 14B Instruct 14.8B	Tight fit	Q4_K_M	8K	26–36
DeepSeek R1 Distill Qwen 14B 14.8B	Tight fit	Q4_K_M	8K	26–36
Qwen2.5 Coder 14B Instruct 128K 14.8B	Tight fit	Q4_K_M	8K	26–36
phi 4 14.7B	Tight fit	Q4_K_M	8K	26–36
gemma 4 12b it 11.9B	Tight fit	Q6_K	8K	21–29
Devstral Small 2 24B Instruct 2512 23.6B	Tight fit	UD-IQ3_XXS	8K	26–35
Mistral Small 3.2 24B Instruct 2506 23.6B	Tight fit	UD-IQ3_XXS	8K	26–35
Magistral Small 2509 23.6B	Tight fit	UD-IQ3_XXS	8K	26–35
Qwen3.8 27B 27.3B	Tight fit	UD-Q2_K_XL	8K	27–36
Qwen2.5 32B Instruct 32.8B	Tight fit	IQ2_XXS	4K	27–37
Qwen2.5 Coder 32B Instruct 32.8B	Tight fit	IQ2_XXS	4K	27–37
Qwen3 32B 32.8B	Tight fit	UD-IQ2_XXS	4K	27–37
Qwen3 VL 32B Instruct 32.8B	Tight fit	UD-IQ2_XXS	4K	27–37
GLM 4 32B 0414 32.6B	Tight fit	UD-IQ2_XXS	8K	28–39
GLM 4.6V Flash 9.4B	Tight fit	Q8_0	8K	27–37
GLM Z1 9B 0414 9.4B	Tight fit	Q8_0	8K	27–37
GLM 4 9B 0414 9.4B	Tight fit	Q8_0	8K	27–37
GLM 4.1V 9B Thinking 9.4B	Tight fit	Q8_0	8K	27–37
gemma 4 31B it 30.7B	Tight fit	UD-IQ2_XXS	8K	17–23
Qwen3.5 9B 9.0B	Tight fit	Q8_0	8K	28–39
MiMo V2.6 Distill Qwen 9B 9.0B	Tight fit	Q8_0	8K	28–39
Ministral 3 8B Instruct 2512 8.5B	Tight fit	Q8_0	8K	27–37
Ministral 3 8B Reasoning 2512 8.5B	Tight fit	Q8_0	8K	27–37
Qwen3.5 27B 26.9B	Runs	UD-IQ2_XXS	8K	30–41
Qwen3.6 27B 26.9B	Tight fit	UD-IQ2_XXS	8K	28–38
Qwen3 VL 8B Instruct 8.2B	Tight fit	Q8_0	8K	28–38
DeepSeek R1 0528 Qwen3 8B 8.2B	Tight fit	Q8_0	8K	28–38
Qwen3 8B 8.2B	Tight fit	Q8_0	8K	28–38
Qwen3 VL 8B Thinking 8.2B	Tight fit	Q8_0	8K	28–38
gemma 3 27b it 27.0B	Runs	UD-IQ2_XXS	8K	23–31
medgemma 27b text it 27.0B	Runs	UD-IQ2_XXS	8K	23–31
Llama 3.1 8B Instruct 8.0B	Runs	UD-Q6_K_XL	8K	33–45
DeepSeek R1 Distill Llama 8B 8.0B	Tight fit	Q8_0	8K	29–39
Apertus 8B Instruct 2509 8.1B	Tight fit	Q8_0	8K	29–39
GLM 4.7 Flash REAP 23B A3B 23.0B	Tight fit	Q3_K_S	8K	155–212
Kwaipilot KAT Coder V2.5 Dev 34.7B	Tight fit	IQ2_XXS	8K	143–195

90 more models in the catalog do not fit this default profile, or ranked below the first 40 that do.

Example: Tight fit fully on your GeForce RTX 4070 Super at Q5_K_S, 4K context, about 25–35 tok/s. Weights 9.56 GB plus 0.63 GB of KV cache — 10.6 GB of 10.9 GB usable. Weights shown in the verdict use GiB (the same 1024³ counting as GPU memory).